Late Endosomes: Sorting and Partitioning in Multivesicular Bodies


  • Robert C. Piper,

    1. Department of Physiology and Biophysics, University of Iowa, Iowa City, IA 52242, USA
    Search for more papers by this author
  • J. Paul Luzio

    Corresponding author
    1. Department of Clinical Biochemistry and Cambridge Institute for Medical Research, University of Cambridge, Addenbrooke's Hospital, Hills Road, Cambridge, CB2 2XY, UK
    Search for more papers by this author

*Corresponding author: J. Paul Luzio, JPL10@CAM.AC.UK


Late endosomes, which have the morphological characteristics of multivesicular bodies, have received relatively little attention in comparison with early endosomes and lysosomes. Recent work in mammalian and yeast cells has given insights into their structure and function, including the generation of their multivesicular morphology. Lipid partitioning to create microdomains enriched in specific lipids is observed in late endosomes, with some lumenal vesicles enriched in lysobisphosphatidic acid and others in phosphatidylinositol 3-phosphate. Sorting of membrane proteins into the lumenal vesicles may occur because of the properties of their trans-membrane domains, or as a result of tagging with ubiquitin. Yeast class E Vps proteins and their mammalian orthologs are the best candidates to make up the protein machinery that controls inward budding, a process that starts in early endosomes. Late endosomes are able to undergo homotypic fusion events and also heterotypic fusion with lysosomes, a process that delivers endocytosed macromolecules for proteolytic degradation.

Characterization and Formation of Late Endosomes

Late endosomes are prelysosomal endocytic organelles defined by the time it takes for endocytosed macromolecules to be delivered to them [reviewed (1,2)]. Although the time of delivery may vary between cell types and between cells in culture and in living tissue, late endosomes are usually loaded 4–30 min after endocytic uptake in mammalian cells. Viewed with the electron microscope, late endosomes are more spherical than early endosomes and are mostly juxtanuclear, being concentrated near the microtubule organizing center. In contrast to early endosomes, they have the appearance of multivesicular bodies (MVBs). They are differentiated from early endosomes by their lower lumenal pH, different protein composition and association with different small GTPases of the rab family [reviewed in (3)] and from lysosomes, because they are enriched in the two mannose 6-phosphate receptors (MPRs), whereas lysosomes have neither (4,5). Characterization of late endosomes in yeast is more problematic than in mammalian cells, but has been made possible with mutants that accumulate late-endosomal compartments, analysis of synchronized endocytosis of marker proteins and dyes, and the recent availability of specific marker proteins (2,6–8)

In the past there has been considerable dispute about whether the transport between early endosomes and late endosomes is best explained by vesicular transport or by the maturation of early endosomes. While both models provide for an intermediate between early and late endosomes, the difference lies in whether the intermediate is a specific transport vesicle budded from the early endosome or is what remains after removal of components from an early endosome (9,10). Late endosomes are also likely to receive traffic directly from the trans-Golgi network (TGN). Aside from their ability to transit to late endosomes by the endocytic route, lysosomal integral membrane glycoproteins (variously referred to as LAMPs, LIMPs or lgps) may be delivered by a more direct route from the TGN to late endocytic compartments, a route that is mediated by the adaptor AP-3 [reviewed in (1,11)]. It remains less clear whether proteins such as the MPRs follow a direct route from the TGN to the late endosome or arrive via early endosomes. MPRs are sorted into clathrin-coated vesicles at the TGN as a result of the interaction of their cytosolic tails with Golgi localized, γ-ear containing ARF-binding proteins (GGAs), rather than by interaction with the clathrin adaptor AP-1 as was previously thought (12–14). Because of the effects of a dominant negative rab 7 mutant (15) and the detection of newly synthesized M6P-tagged proteins within early endosomes (16), the MPRs may first transit to early endosomes rather than directly to late endosomes. Elegant experiments in yeast, however, suggest that GGAs control formation of TGN-derived vesicles destined for delivery to late endosomes (8).

Fusion, Fate and Fission

It is not only the biogenesis of late endosomes which has received much attention in recent years, but also their fate. It is clear that late endosomes can participate in homotypic fusion reactions with themselves (17) and heterotypic fusions with lysosomes (1). The former may play an important role in morphological remodeling and the latter in creating a hybrid organelle that has been proposed to represent ‘the cell stomach’ (18), where degradation occurs. Lysosomes are reformed from the hybrid organelle by a process requiring condensation of lumenal content (19) and probably vesicular removal of membrane components that are absent from, or present at reduced concentration in, lysosomes. The reformation of lysosomes from hybrid organelles is a maturation process, such that in the early stages the organelles contain MPRs and therefore fall into the general classification of endosomes, whereas later they reach a point where they no longer have MPRs and can be classified as lysosomes. The existence of hybrid organelles with the characteristics of both conventional late endosomes and lysosomes may go some way to explaining apparent differences in data from different laboratories on the site of accumulation of cholesterol in cells from patients with Niemann–Pick type C disease, lacking the multispanning membrane protein NPC1, or in normal fibroblasts treated with the hydrophobic amine U18666A (20,21). The effects of this drug can be rescued by overproducing NPC1 (22), implying that NPC1 is the target of U18666A or that the two act along the same pathway. If the hybrid organelle is a major site of low-density lipoprotein (LDL) degradation and cholesterol accumulation under these conditions, its classification as endosome or lysosome will depend solely on the stage of maturation to lysosomes and the consequent presence or absence of MPRs. Using cell-free systems, great strides have been made in understanding the mechanisms of fusion events undertaken by late endosomes. In cell-free assays, homotypic late-endosome fusion has been shown to be inhibited by antibodies to syntaxin 7, syntaxin 8, Vti1b and VAMP8 (endobrevin) (17). In contrast, whereas antibodies to syntaxin 7 also inhibit heterotypic late endosome–lysosome fusion, antibodies to VAMP8 do not (23), suggesting the possibility that an alternative v-SNARE is required (Figure 1). The complexity of the data concerning which SNAREs are required for which fusion events in the endocytic pathway may be explained if combinatorial interactions (24) are allowed, e.g. between one v-SNARE and more than one t-SNARE and/or vice versa. A further contribution to the complexity may be that different heterotrimeric t-SNARE complexes (25) have some common individual polypeptide chains. We have obtained evidence consistent with combinatorial interactions involving syntaxin 7 occurring in vivo, from a co-immunoprecipitation study in B16 melanoma cells. In these cells antibodies to syntaxin 7 coprecipitated Vti1b, syntaxin 6, VAMP8 and VAMP7 (also known as TI-VAMP) (26). VAMP7 has been implicated previously as a SNARE required late in the endocytic pathway (27), suggesting that it and VAMP8 might be alternative v-SNAREs (R-SNAREs) interacting with t-SNARE (Q-SNARE) complexes containing syntaxin 7.

Figure 1.

Anatomy and trafficking pathways of the late endosome. (A) Distribution of several proteins from mammalian cells or yeast in the late endosome or in yeast vacuoles that reflect the enrichment of proteins on lumenal vesicles or the limiting membrane in the MVB. (B) The proteins that may control several trafficking pathways relevant to the late endosome have now been identified. These machinery sets control the fusion of late endosomes, the formation of internal vesicles that accumulate in the late endosome, and recycling out of the late endosome. Shown where possible are both the putative yeast and mammalian orthologs. The molecular identity of each component is given below by its gene name and GenBank accession number: VAM2/VPS41/CVT8 (gb: CAA86802), huVPS41 (gb: NP_055211), VAM6/VPS39/CVT4 (gb: CAA98643), huVps39 (gb: AB018313), PEP3/VPS18/VAM8 (gb: CAA97720), huVPS18 (gb: AAG34679), PEP5/END1/VPS11 (gb:CAA90202), huVPS11 (gb: AAG34677), VAM3/PTH1 (gb: CAA99304), Syntaxin7 (gb: NP_058077), NYV1 (gb: CAA97654), Vamp8/Vamp7? (gb: NP_005629/gb: NP_003752), YPT7 (gb: CAA88515), Rab7/Rab9? (gb: P18067/gb: XP_010326), VPS16/VAM9 (gb: AAB68176), huVPS16 (gb: NP_072097), VPS33/SLP1/PEP14 (gb: AAB82354), rVps33a/Vps33b? (gb: AAC52985/gb: AAC52986), AUT7/APG8 (gb: CAA84899), GATE-16 (gb: 10835904), CCZ1 (gb: CAA85088), VAC8 (gb: AAB64490), VPS4/END13(gb: AAB68107), VPS4a (gb: NP_037377),SKD/Vps4b (gb: P46467), VPS20 (gb: CAA89223), STP22/VPS23 (gb: CAA42351), TSG101(gb: 1772664), VPS24/DID3 (gb: CAA81876), VPS25 (gb: CAA89632), huVPS25 (gb: NP_115729), VPS27/DID7 (gb: CAA96282), Hrs1, Hrs2 (gb: NP_032270), VPS28 (gb: AAB40936), huVPS28 (gb: NP_057292), SNF7/VPS32/DID1, VPS36 (gb: AAB67493), CGI-145 (gb: NP_057159), MOS10/VPS60 (gb: AAB64922), DID2 (gb: NP_012961), DID4/VPS2/REN1 (gb: NP_012924), BC-2 (gb: AAC00005), YHL002w (gb: AAB68427), VPS5/GRD2/PEP10 (gb: AAB62976), SNX1 (gb: G02522), VPS17/PEP21 (gb: CAA99331), PEP8/VPS26/GRD6 (gb: CAA89344) huVPS26 (gb: XP_005979), VPS29/PEP11 (gb: AAB68947), huVPS29 (gb: AAF89952), VPS35/GRD9 (gb: CAA89449), huVPS35 (gb: AAF89953), PACS 1a, 1b (gb: AAC31815, AAC31816), TIP47(gb:), LYST (gb: XP_002041) rab9 (gb: XP_010326).

Recycling of proteins out of late endosomes is important for allowing certain components to avoid degradation in the lysosome and to serve as part of the mechanism to reform lysosomes and late endosomes from hybrid organelles. Both MPRs and furin are thought to recycle to the TGN from late endosomes (28,29). Whilst some proteins interacting with the cytosolic tails of these membrane proteins have been shown to be required for retrieval (30,31), the identity of the coat proteins involved is not clear (Figure 1). Components of the retromer complex identified as responsible for retrograde traffic from the prelysosomal/endosomal compartment to the trans-Golgi in yeast have orthologs in mammalian cells which are obvious candidates to be involved (32–34).

Lipid Partitioning in Late Endosomes: The Origin of Multivesicular Bodies

One of the hallmarks of late endosomes is the presence of internal membranes (Figure 1). Hence, the reason that late endosomes have also been referred to as multivesicular bodies (MVBs) or multivesicular endosomes (MVEs). The accumulation of internal membranes starts at the early endosome and is thought to continue as the endosome ‘matures’ to a late endosome. The lipid composition of late endosomes differs from that of earlier endocytic compartments, being enriched in triglycerides, cholesterol esters and selected phospholipids including lysobisphosphatidic acid (LBPA) (35). Compared to the limiting membrane, the internal membranes of mammalian late endosomes/MVBs are heavily enriched in either LBPA (36) or phosphatidylinositol 3-phosphate (PI3P) (37), the latter being in distinct regions low in LBPA. LBPA is predicted to be highly hydrophobic, cone shaped and to bend membranes, implying a possible function in formation of MVBs. Intralumenal vesicles can also be observed in the vacuoles of Saccharomyces cerevisiae that do not contain active vacuolar proteases. Like late endosomes from mammalian cells, yeast vacuoles also have a distinct membrane composition based on the trafficking of lipid dyes such as FM4-64 and NBD-PC to the limiting and internal membranes, respectively (38,39). Similarly, PI3P has been found on the intralumenal vesicles of vacuoles in yeast (37). The finding that in yeast, generation of phosphatidyl-3,5-bisphosphate from PI3P by the action of Fab1p is required for the formation of MVBs (40,41) also supports a role for lipid sorting in MVB formation and function, as does the observation that in Mel JuSo cells, a human melanoma cell line, the PI3-kinase inhibitor wortmannin prevents the biogenesis of MVBs (42). However, the relationship between the lipids of the limiting and internal membranes of MVBs remains poorly understood. In addition, in normal rat kidney (NRK) fibroblasts, the swollen endosome phenotype induced by wortmannin appears to be mainly a consequence of endocytic membrane influx coupled with the failure to recycle membrane to other cellular destinations and not the inhibition of MVB biogenesis (43). Recent work on cells from patients with Niemann–Pick type C disease, that lack the multispanning membrane protein NPC1, and on U18666A-treated fibroblasts has shown that cholesterol accumulation in late endocytic compartments is accompanied by sequestration of sphingolipids, multilamellar body formation and the accumulation of low-density, detergent-resistant membranes or rafts (44), thus implying a central role for removal of cholesterol in the normal morphology and function of late endosomes. NPC1 is clearly implicated in trafficking of cholesterol out of late endosomes by a vesicular transport mechanism and has been suggested to be required for retrieval of membrane constituents from the lumenal vesicles of MVBs and their assembly into transport vesicles (45), budding of vesicles from late endocytic organelles and/or control of organelle partitioning events (22).

Why Localize Proteins to Internal Membranes?

Sorting to the lumenal compartment of endosomes has long been recognized as a means of degradation for receptors such as the EGFR and provides a means of degrading both lumenal and cytosolic domains of integral membrane proteins (46). Similarly, in yeast a considerable number of integral membrane proteins are degraded by the vacuole (47,48) and those that have been specifically localized, traffic to the vacuole interior (49–52). Wild-type yeast, however, lack lumenal vesicles in endocytic compartments. Vesicles can only be visualized in mutants that lack active vacuolar proteases (pep4Δ). This may be due to technical problems associated with preservation and/or identification of the late-endosomal compartment by EM or due to a very robust lipase(s) used for the degradation of the lipid vesicles.

Apart from the degradative fate of proteins sent to the endosomal lumen, a select set of proteins in mammalian cells appear quite stable in internal vesicles. These proteins include a set of tetraspanins (including CD63/LAMP3, CD37, CD81 and CD82) and the MHC class II complex which is found in the late-endosome-like MHC-class II compartment (MIIC) (53,54). The localization of these proteins to lumenal vesicles provides a very important immune function in antigen-presenting cells, since upon stimulation these late-endosomal compartments fuse to the plasma membrane and release the intralumenal vesicles. These released vesicles, termed ‘exosomes’, have been shown to promote B and T lymphocyte activation (55). Interestingly, in B cells lacking CHS (Chediak–Higashi syndrome/LYST/Beige protein) the internal membranes of MVEs contain very little MHC class II or CD63, even though these MVEs contain lumenal vesicles that are exocytosed after stimulation (56). All the proteins of the exosomes are also present on the cell surface. The fact that CD63 is efficiently endocytosed and delivered to late endosomes and that it is highly enriched in internal vesicles brings into question the source of the CD63 that continually populates the plasma membrane. Is this CD63 from a separate pool than that in internal vesicles or can the CD63 already in internal vesicles be recruited back to the plasma membrane, possibly after fusion with the limiting membrane of the late endosome? Similar experimental approaches have provided evidence for both these models; more studies are therefore needed to resolve the issue (57,58).

Why Are MPRs Enriched on the Internal Membranes of Late Endosomes?

Multilamellar membranes within late endosomes are frequently associated with MPR labeling, while the limiting membrane of these compartments often remains unlabeled (4,5). This apparently paradoxical steady-state distribution may be accounted for, when one considers that peripheral receptors are likely to be rapidly retrieved to the TGN to collect more mannose-6-phosphate-tagged cargo for delivery to late endosomes, whilst receptors that become associated with the internal membranes are inaccessible to the cytosolic retrieval machinery. Consistent with this interpretation, Hirst et al. (28) have demonstrated that in Hep-2 cells, MPRs are rapidly retrieved from the periphery of multivesicular endosomes. However, antibodies against LBPA, when internalized from the extracellular medium prevent MPR trafficking out of late endosomes and cause cholesterol accumulation within these organelles (20). Further evidence for a role for cholesterol in the trafficking of MPRs out of late endosomes, and indeed in the selective sorting of MPR to lumenal vesicles, has come from studies of the Chinese hamster ovary cell mutant LEX2 (Lysosome-Endosome X2, a mutant defective in the degradation of endocytosed low-density lipoprotein), which accumulates MPRs in arrested MVBs, a phenotype that can be reversed by restoration of normal cellular free cholesterol levels (59). Thus, it is possible that in late endosomes a dynamic equilibrium exists between the lumenal vesicles and the limiting membrane, and that MPRs sorted into the lumenal vesicles may not be irreversibly destined for degradation.

How Are Proteins Targeted to the Endosome Lumen?

It is critically important to distinguish between targeting to late endosomes versus into late endosomes. In mammalian cells, the targeting motif in the cytosolic tail of CD63 that sorts it to late endosomes and lysosomes is a tyrosine-based sorting motif similar to that which functions in the sorting of Lamp1, Lamp2 and HLA-DM (60). However, CD63 is enriched on internal membranes while Lamp1/2 and HLA-DM are not (Figure 1). Thus, there must be other requirements for sorting into intralumenal vesicles.

Studies in yeast have provided two models for how intralumenal sorting happens. The first is that the composition of transmembrane domains can determine sorting into lumenal membranes. Trans-membrane domains that contain polar amino acids were able to confer lumenal sorting on a variety of type II GFP-tagged membrane proteins (61). In view of the observation that the lumenal membranes of mammalian late endosome and yeast pep4Δ vacuoles have a different lipid composition than the limiting membrane, these data provide a good model for how the differential solubility of trans-membrane domains could control partitioning into regions of the endosome surface that would later undergo involution into the lumen. Such a model cannot account for the sorting of all proteins into lumenal membranes, since some proteins that are sorted to the lumen also have the capacity to recycle back to the cell surface under some conditions (62,63). Thus, these proteins could somehow be tagged for sorting into the lumen for degradation.

Ubiquitin appears to be such a tag for lumenal sorting. It is well established in both yeast and mammalian cells that ubiquitination of some cell surface proteins can cause their internalization from the cell surface (47,48). However, it is less clear as to whether the correlation between ubiquitination and down-regulation for a variety of mammalian cell surface proteins is due to increased internalization, since many studies have not examined this directly. It now appears that ubiquitin can also mediate specific sorting into the yeast vacuole, since tagging proteins that normally recycle between the TGN and endosomes, or proteins that normally reside at the limiting membrane of the yeast vacuole, results in their localization to the vacuole interior (64). This effect was observed using a mutant ubiquitin moiety unable to form polyubiquitin chains via Lys29, Lys48 or Lys63, suggesting that monoubiquitination is sufficient for lumenal sorting. Furthermore, in yeast mutants deficient for ubiquitination, Ste6p, which normally undergoes intralumenal vacuolar degradation, accumulates on the limiting membrane of the vacuole (65). The observation that ubiquitin acts as a dominant sorting signal for protein sorting to lumenal membranes offers a mechanism by which cell surface receptors can be diverted from the recycling pathway to the degradative pathway in a regulated manner. Corroborating evidence for this model in mammalian cells is provided by studies of EGFR and its ubiquitination and accelerated degradation by the E3 ligase c-cbl (66–68). Upon tyrosine kinase activation, c-cbl associates with EGFR and mediates its ubiquitination, most likely in the early endosome (67,69). Overexpression of c-cbl greatly enhances the level of EGFR ubiquitination and its rate of ligand-dependent degradation without altering the rate at which it is internalized from the cell surface (66). C-cbl mutants that lack the ability to ubiquitinate EGFR do not enhance EGFR degradation, suggesting that ubiquitination of EGFR via c-cbl enhances the entry of EGFR into the lumenal degradative pathway which occurs at a postinternalization step (68). Interestingly, c-cbl also promotes the lysosomal degradation of the EGFR-associated cytosolic signaling subunits Shc and Grb2 which are known to stay associated with EGFR throughout its transit along the endocytic pathway (70–72), thereby enhancing the mechanism by which signal transduction is down-regulated by sorting to lysosomes. A dileucine motif prior to the kinase domain in EGFR (73) also appears to confer sorting into lumenal membranes. It has been suggested that the motif in ubiquitin that serves as a cell surface internalization signal is also a dileucine (74). Thus, it will be interesting to determine whether the dileucine motif of EGFR works directly, possibly by mimicking the sorting information of ubiquitin, or whether the presence of the dileucine motif works indirectly to promote ubiquitination of EGFR. Recent studies with the IL-2 receptor suggest the model that ubiquitin acts directly as a sorting tag for the degradative compartment. The IL-2 beta chain is ubiquitinated and also undergoes degradation in the lysosome. Mutation of the lysines in the cytosolic tail that undergo ubiquitination dramatically stabilizes the IL-2R without affecting its rate of internalization. Furthermore, this mutant fails to be efficiently transported to Lamp1-positive compartments and instead is localized to TfR-positive endosomes (75).

What is the Machinery that Controls Inward Budding?

The best candidates for proteins that promote the formation of lumenal membranes are encoded by the class E VPS genes originally described in yeast. Yeast mutants lacking any of the class E genes accumulate an exaggerated prevacuolar/late-endosome compartment that accumulates both recycling TGN proteins and vacuolar proteins (76). Loss of class E protein function rapidly leads to the accumulation of large multilamellar cisternal compartments near the vacuole (38,76,77). Although these mutants were originally defined as defective in their ability to sort proteins out of the late-endosomal compartment, resulting in a kinetic delay out of late endosomes (78,79), all the data so far suggest that the primary defect is the inability to promote sorting into the lumenal degradative compartment. Accordingly, class E mutants stabilize proteins that would otherwise occupy the vacuolar lumen, on the limiting membrane (40). By electron microscopy of whole cells, class E vps mutant cells show no internal vacuolar vesicles and do not accumulate lumenal PI3P (37). Class E mutants do not show defects in internalization from the plasma membrane (80); however, they have been identified in screens for mutants that cause accumulation at the plasma membrane of proteins that normally undergo PEP4-dependent degradation within the vacuole (49,81,82). This is consistent with the loss of sorting into a degradative pathway, thus allowing endocytosed proteins to be recycled back to the cell surface (63). Finally, class E mutations have been found to suppress the loss of free ubiquitin that results from loss of DOA4 (83). Doa4p is a deubiquitinating enzyme that is required for the efficient removal of ubiquitin from proteins prior to their sorting into the vacuolar lumen for degradation (84). Doa4p itself is recruited to endosomes in class E vps mutants and in doa4Δ mutants, enough ubiquitin is degraded within the vacuole so that doa4Δ mutants show a significant growth defect. Class E vps mutations are able to restore ubiquitin levels by stabilizing ubiquitinated proteins on the limiting membrane of endosomes, preventing the entry of polyubiquitin chains into the degradative pathway. Many of the class E proteins from both yeast and mammalian cells have now been identified, which allows some of the pieces of the puzzle to fall into place (Figure 1B). What do we know about the pieces themselves?

Vps4p assembles into multimers and contains both an N-terminal coiled-coil domain and a AAA ATPase domain (85). Based on the observation that other AAA ATPases, such as NEM-sensitive factor (NSF), work as protein-specific chaperones, it has been proposed that Vps4p catalyzes the disassembly of specific protein complexes, possibly other class E proteins (85). Mutations in the ATPase domain that block ATP hydrolysis or ATP binding have been identified and convert Vps4p into a dominant-negative mutant (86) that is tightly associated with membranes in a large Triton-insoluble complex (85). Presumably, the dominant-negative effect is caused by the poisoning of endogenous Vps4p oligomers or the titration of specific Vps4p substrate proteins. Expression of mammalian Vps4p orthologs containing the same dominant-negative mutations results in swollen endosomal compartments that accumulate TfR, Lamp1, endocytosed dextran and HRP as well as other endocytic markers such as EEA1, consistent with the class E morphology in yeast vps4Δ mutants (87,88). Electron microscopy of these cells shows the accumulation of large vacuolar structures reminiscent of early sorting endosomes that are continuous with a network of tubular vesicular structures. At odds with the proposed role for Vps4p in the formation of lumenal vesicles was the observation that the vacuolar structures contained some internal membranes (88). However, it is possible that only a partial block in Vps4p function was effected during short-term expression of the dominant-negative allele.

Many of the other class E Vps proteins belong to a family of small coiled-coil proteins that includes Snf7p, Mos10p, Vps20p, Vps24, Did2p and Did4p (82,84,85). Despite the high level of identity amongst some of these family members, none of these is redundant, and only in the case of Vps24p and highly related Did4p could one partially suppress the defects of another when over-expressed (84). There are subtle differences between the corresponding mutant phenotypes, suggesting some functional differences between these proteins (82,84). Nonetheless, the presence of coiled-coil domains within these proteins suggests that they may complex together to form a higher ordered structure that could serve as a substrate/receptor for Vps4p. Several of these family members appear to interact with other class E Vps proteins. Vps20p interacts with both Vps25p and Vps36p (89). Snf7p interacts with Vps4p by yeast 2-hybrid and Snf7 together with Vps24 assembles in a large pelletable membrane-bound complex in the absence of Vps4p (85,89).

Vps23p and Vps28p have been shown to interact, as have their respective mammalian orthologs, TSG101 and mVps28p (90,91). In the absence of TSG101, internalized EGFR escapes degradation and is recycled to the plasma membrane, and similar recycling of Ste2p is observed in yeast cells lacking VPS23 (81,91). These results are consistent with the inability of receptors to enter the lumenal degradative pathway. Both Vps23p and TSG101 have been noted previously to contain a motif similar to the ubiquitin-binding region of E2 conjugation enzymes within the N-terminal region, while the C-terminal region confers the Vps28p binding (90–92). Given the role of ubiquitin in sorting proteins into the lumenal degradative compartment, it may be that Vps23p acts as a specific sorting receptor that recognizes ubiquitinated proteins.

Vps27p has been used as a founding member to define two of its modular domains, a FYVE RING finger domain that binds to PI-3P as well as an N-terminal VHS domain. Vps27p appears to be the yeast ortholog of mammalian Hrs (recent sequencing analysis has revealed that Hrs-(I) and Hrs-2 are actually one and the same) (93). Cultured cells from mouse embryos lacking Hrs accumulate enlarged early endosomal compartments, consistent with defects observed in cells expressing dominant-negative Vps4. Hrs is tyrosine phosphorylated by a number of receptor tyrosine kinases (RTKs) and this phosphorylation event is likely to occur at early endosome compartments (94). Clues as to what Hrs may be doing come from studies that have identified a plethora of interacting proteins. The coiled-coil domain of Hrs mediates interaction with Snx1, which plays a role in the EGF receptor degradation (95) and two highly related molecules, STAM and Hbp (93). Both STAM and Hbp are related to the chicken protein EAST and all contain an N-terminal VHS domain, an SH3 domain, and a TAM domain. STAM and Hbp also become tyrosine phosphorylated by RTKs and have been implicated in potentiating signal transduction through the Jak/Stat pathway and the control of platelet derived growth factor (PDGF) receptor degradation, respectively (93). Mice lacking STAM are relatively healthy; however, the lack of phenotypic similarity of STAM knockout mice with Hrs –/– mice is likely due to the redundancy of STAM, Hbp, and another related molecule STAM2 (96,97). The yeast genome encodes one STAM-related protein that has both an SH3 domain and a VHS domain but lacks a TAM domain (YHL002w). Cells lacking YHL002w display a class E vps defect (Urbanowski and Piper, unpublished observation), suggesting that the STAM/Hbp/EAST family of proteins may play a role in MVB formation. Interestingly, the SH3 domain of Hbp interacts with the deubiquitinating enzyme UBPY and may provide the means to recruit deubiquitinating activity to the endosome similar to the recruitment of Doa4p to endosomal compartments to effect deubiquitination of proteins destined for degradation within the vacuolar lumen (84). Hrs has also been shown to bind to Eps15 within the region required for binding to the AP-2 clathrin adaptor complex (98). Accordingly, overexpression of Hrs inhibits transferrin uptake, suggesting a role for Hrs in clathrin-mediated internalization, in addition to an additional role in effecting calcium-stimulated exocytosis via interaction with SNAP-25 (99). Alternatively, the interaction of Hrs with Eps15 may indicate that some of the clathrin-budding machinery may also participate in sorting events at the early endosome.

While much of the current work on the above proteins is focused on identifying all the key players and their physical interactions, it is still unclear what these proteins are actually doing and how they contribute to the formation of internal vesicles. One possibility is that they form a multisubunit membrane coat that works to involute membranes towards the lumen. However, none of the class E Vps proteins has been found within lumenal membranes or to undergo PEP4-dependent degradation and thus they are not included in the lumenal vesicle. Alternatively, they may act to sort and segregate proteins designated for degradation from those destined to recycle to earlier compartments. Their absence could result in disorganized subdomains of the limiting membrane, not only producing a failure of budding into the vacuole, but also trapping recycling proteins at the endosome. Finally, the class E proteins may direct the specific sorting or synthesis of particular lipids on the cytosolic face to form a subdomain conducive to negative curvature and involution (see above).

Potential for Regulated Sorting into Lumenal Vesicles of the MVB

The ability to define a separate sorting step for entry into the lumen of the MVB allows the possibility that this sorting step could be acutely regulated and contribute to the overall effect of a variety of signal transduction pathways by regulating degradation of RTKs. Likewise, activation of RTKs themselves may influence the rate at which lumenal sorting occurs. For instance, both Hrs and its partner Hbp are phosphorylated by RTKs. This results in shunting Hrs off the early endosomal membranes (94), which would alter its function. Regulation of RhoB is another means by which this pathway could be regulated. RhoB localizes to early endosomes and MVBs, and over-expression of WT or constitutively active RhoB slows EGFR degradation and its delivery to lysosomes (100). Other molecules could act as specific adaptor-type molecules that could efficiently direct other proteins into the lumenal degradative compartment. Candidates for this function include the tetraspanin CD82, which acts as a tumor suppressor. CD82 binds to the EGFR and accelerates ligand-dependent degradation of EGFR (101). Since CD82 occupies ‘exosomes’, it could help divert EGFR to the lumenal compartment (54). Another candidate is the E3-13.7 protein of adenovirus that promotes degradation of EGFR in the absence of ligand. E3-13.7 is localized to early endosomes and MVBs, where it is found in a complex with internalized EGFR (102). In cells expressing E3-13.7, the EGFR accumulates in MVBs and, since the receptor internalization rate is the same in the absence or presence of E3-13.7, the likely site of action of E3-13.7 is to promote sorting of EGFR into the lumenal degradative pathway. Although the degradation of EGFR is unlikely to be mediated by c-cbl-dependent ubiquitination of EGFR, since E3-13.7-dependent degradation occurs without activation of the tyrosine kinase domain, it will be interesting to determine whether E3-13.7 promotes ubiquitination of EGFR by some other mechanism or whether it plays a more general role in the trafficking of other molecules within the endosome. Thus, there may be both generalized ways to up-regulate protein sorting into the lumenal degradative pathway as well as ways to modulate the entry of specific proteins into the degradative pathway. It will be interesting to determine how a variety of factors (such as the nef protein of HIV) might influence this sorting step (103).


We thank Barbara Mullock and Gudrun Ihrke for helpful comments and suggestions. Experimental work in our laboratories is funded by grants from the National Institutes of Health to RCP (GM58202), from the Medical Research Council and the Wellcome Trust to JPL, and from the Human Frontiers Science Program to RCP, JPL and David E. James of the University of Queensland, Australia.